EUVIMEDEUVIMED entdecken
Open Access · CC BY-NC 4.0

Estimation of Potential Fishing Grounds Using the Random Forest Machine Learning Algorithm with MODIS Satellite Data

Hasbi Husaini, Ahmad Mujab Arba'i, Muhammad Rajaskana Syahputra

YKP JOURNAL · 2026

DOI: 10.63639/ykp.v1i4.15 ↗

Kurz gefasst

Fachartikel in der Zeitschrift YKP JOURNAL, 2026. Verfasst von Hasbi Husaini und 2 weiteren Autor:innen. Der Volltext ist frei zugänglich (Open Access, CC BY-NC 4.0).

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

North Sumatra Province, located within Indonesia’s Fisheries Management Area (WPP) 571, has substantial marine capture fisheries potential along its eastern coastline. However, increasing fishing intensity has led to higher fuel consumption and carbon emissions from fishing vessels, emphasizing the need for spatially efficient fishing strategies. This study aims to estimate and validate potential fishing zones (PFZ) using a Random Forest machine learning algorithm integrated with MODIS-Aqua satellite data to enhance operational efficiency and support emission mitigation. Four oceanographic parameters were employed as predictor variables: Sea Surface Temperature (SST), chlorophyll-a concentration (Chl-a), normalized fluorescence line height (nFLH), and particulate organic carbon (POC), covering the period 2013–2020. Ecological threshold values were applied to generate PFZ and Non-PFZ labels, and the dataset was divided into 80% training and 20% testing subsets using stratified random sampling. Model performance was evaluated using confusion matrix analysis, precision, recall, F1-score, and Cohen’s Kappa coefficient. Results indicate that all variables were normally distributed and significantly correlated (R > 0.5; R² > 0.7). The Random Forest model achieved consistently high predictive accuracy, exceeding 90% across all evaluated years. Compared to a conventional threshold-based method, the machine learning approach produced more stable and spatially consistent PFZ delineations by capturing nonlinear environmental interactions. This study demonstrates that integrating satellite oceanography and machine learning provides a robust, data-driven framework for improving fishing efficiency while contributing to carbon emission reduction in capture fisheries management.

Zitieren

APA

Husaini, H., Arba'i, A. M., & Syahputra, M. R. (2026). Estimation of Potential Fishing Grounds Using the Random Forest Machine Learning Algorithm with MODIS Satellite Data. YKP JOURNAL. https://doi.org/10.63639/ykp.v1i4.15

Vancouver

Husaini H, Arba'i AM, Syahputra MR. Estimation of Potential Fishing Grounds Using the Random Forest Machine Learning Algorithm with MODIS Satellite Data. YKP JOURNAL. 2026. doi:10.63639/ykp.v1i4.15

BibTeX
@article{Husaini2026,
  author = {Husaini, Hasbi and Arba'i, Ahmad Mujab and Syahputra, Muhammad Rajaskana},
  title = {Estimation of Potential Fishing Grounds Using the Random Forest Machine Learning Algorithm with MODIS Satellite Data},
  journal = {YKP JOURNAL},
  year = {2026},
  doi = {10.63639/ykp.v1i4.15}
}
Kostenlos starten →